Skip to content

Latest commit

 

History

History
510 lines (384 loc) · 14.1 KB

File metadata and controls

510 lines (384 loc) · 14.1 KB

NotebookMLX User Guide

Welcome to NotebookMLX! This comprehensive guide will help you get started with converting PDFs into engaging conversational podcasts using Apple's MLX framework.

Table of Contents

  1. Getting Started
  2. Understanding the Interface
  3. Working with Sources
  4. Chat Interactions
  5. Studio Features
  6. Exporting Content
  7. Troubleshooting
  8. Tips and Best Practices

Getting Started

First Launch

When you first open NotebookMLX, you'll see the main interface with three primary areas:

  1. Sources Panel (left) - Manage your PDF documents
  2. Chat Panel (center) - Interact with your documents
  3. Studio Panel (center, alternative view) - Create podcasts and mind maps

Quick Start Tutorial

Step 1: Upload Your First Document

  1. Click the "Add Source" button in the Sources panel
  2. Select a PDF file from your computer (max 200MB)
  3. Wait for the document to be processed (this may take a few minutes)
  4. The document will appear in your Sources panel with a "Completed" status

Step 2: Start a Conversation

  1. Select your uploaded document by checking the box next to it
  2. Type a question in the chat input at the bottom
  3. Press Enter or click Send
  4. NotebookMLX will analyze your document and provide a response with citations

Step 3: Generate Your First Podcast

  1. Switch to the Studio panel by clicking the "Studio" tab
  2. Ensure your document is selected
  3. Choose voices for Speaker 1 and Speaker 2
  4. Click "Generate Podcast"
  5. Wait for the generation process to complete
  6. Listen to your generated podcast!

Understanding the Interface

Main Navigation

The application uses a tab-based interface:

  • Chat Tab: Interactive Q&A with your documents
  • Studio Tab: Content creation tools (podcasts, mind maps, videos)

Sources Panel Features

  • File Upload: Drag and drop or click to upload PDFs
  • Source Selection: Check boxes to select documents for chat/generation
  • Status Indicators:
    • 🟡 Processing - Document is being analyzed
    • 🟢 Completed - Ready for use
    • 🔴 Error - Processing failed
  • Search: Filter sources by filename
  • Refresh: Update source status

Status Indicators

Throughout the interface, you'll see various status indicators:

  • Loading spinners: Operations in progress
  • Progress bars: Show completion percentage for long tasks
  • Color-coded badges: Quick status identification
  • Toast notifications: Success/error messages

Working with Sources

Supported File Types

NotebookMLX currently supports:

  • PDF documents (.pdf)
  • Plain text files (.txt)

File Size Limits

  • Standard upload: Up to 200MB per file
  • Large files: Automatically use chunked upload for files ≥8MB
  • Total content: Up to 100,000 characters extracted per document

Document Processing

When you upload a document, NotebookMLX:

  1. Extracts text from your PDF using advanced parsing
  2. Chunks content into manageable sections (1000 words each)
  3. Cleans and preprocesses text using Qwen2.5-1.5B model
  4. Indexes content for fast retrieval during chat and generation

Managing Sources

Selecting Sources:

  • Check individual sources for targeted analysis
  • Use "Select All" to include all completed sources
  • Selected sources appear highlighted in blue

Removing Sources:

  • Click the trash icon next to any source
  • Confirm deletion in the popup dialog
  • This permanently removes the source and its processed data

Refreshing Sources:

  • Click the refresh icon to update source statuses
  • Useful if processing seems stuck or to check for updates

Chat Interactions

Starting a Conversation

  1. Select Sources: Choose one or more processed documents
  2. Ask Questions: Type natural language questions about your content
  3. Review Responses: Get detailed answers with source citations
  4. Follow Up: Continue the conversation with related questions

Types of Questions You Can Ask

Summarization:

  • "What are the main findings in this research paper?"
  • "Summarize the key points of this document"
  • "What are the conclusions?"

Analysis:

  • "Compare the methodologies discussed"
  • "What are the strengths and weaknesses mentioned?"
  • "How do these results relate to previous research?"

Specific Information:

  • "What does the paper say about machine learning applications?"
  • "Find information about data collection methods"
  • "What are the limitations mentioned?"

Creative Interpretation:

  • "Explain this concept as if I'm a beginner"
  • "What are the practical implications?"
  • "How could this be applied in industry?"

Understanding Citations

Each response includes citations that show:

  • Source Document: Which PDF the information came from
  • Relevance Score: How closely the source matches your question (0-100%)
  • Clickable Links: Click citations to see the source context

Chat Management

Clearing Chat History:

  • Click the "Clear Chat" button to start fresh
  • This only clears the conversation, not your sources

Exporting Conversations:

  • Use the Export button to save chat history
  • Available formats: PDF, HTML, Markdown, JSON

Studio Features

The Studio is where you create engaging content from your documents.

Podcast Generation

Basic Podcast Creation:

  1. Select Sources: Choose the documents you want to discuss
  2. Configure Speakers:
    • Speaker 1: Usually the "teacher" or expert
    • Speaker 2: The "curious learner" asking questions
  3. Choose Voices: Select from built-in or custom trained voices
  4. Generation Options:
    • Enable "Dramatic Enhancement" for more engaging conversation
    • Set target duration (optional)
    • Choose conversation style

Voice Options:

  • Built-in Voices:

    • Default Male/Female voices
    • Multiple language options
    • Ready to use immediately
  • Custom Voices:

    • Train your own voices with audio samples
    • Minimum 2-3 minutes of clear audio needed
    • Process takes 5-15 minutes depending on quality settings

Podcast Generation Process:

  1. Transcript Generation (2-5 minutes): Creates conversation script
  2. Dramatic Enhancement (1-3 minutes): Makes dialogue more engaging
  3. Audio Synthesis (5-15 minutes): Converts text to speech
  4. Final Assembly (1-2 minutes): Combines all audio segments

Mind Map Creation

Generating Mind Maps:

  1. Select your source documents
  2. Click "Generate Mind Map" in the Mind Map Studio
  3. Wait for processing (1-3 minutes)
  4. Interact with the generated visualization

Mind Map Features:

  • Interactive Nodes: Click to explore topics
  • Hierarchical Structure: Central topic with branching subtopics
  • Source Linking: See which documents contributed to each node
  • Export Options: Save as image or interactive HTML

Video Generation

Creating Videos from Podcasts:

  1. First generate a podcast (see above)
  2. Switch to Video Studio
  3. Select your completed podcast
  4. Choose video template:
    • Simple: Static background with title
    • Waveform: Animated audio visualization
    • Transcript: Synchronized text display

Video Customization:

  • Background colors/images
  • Text overlay options
  • Font and styling choices
  • Transition effects

Voice Training

Training Custom Voices:

  1. Go to Voice Studio
  2. Click "Train New Voice"
  3. Upload 3-10 audio files (WAV format recommended)
  4. Provide voice name and description
  5. Choose quality setting:
    • Fast: 2-5 minutes, good quality
    • Balanced: 5-10 minutes, better quality
    • High: 10-20 minutes, best quality

Audio Requirements for Training:

  • Clear speech without background noise
  • Multiple sentences/phrases
  • Consistent speaking pace and tone
  • Total duration: 2-10 minutes of audio

Exporting Content

Export Modal Overview

The Export Modal provides unified access to all export options:

  1. Click any "Export" button in the interface
  2. Choose your content type (Chat, Podcast, etc.)
  3. Select export format
  4. Configure options
  5. Download your file

Chat Exports

PDF Export:

  • Professional document format
  • Includes conversation history
  • Optional cover image
  • Formatted for printing

HTML Export:

  • Interactive web page
  • Clickable citations
  • Responsive design
  • Easy sharing

Markdown Export:

  • Plain text format
  • Great for documentation
  • Version control friendly
  • Cross-platform compatible

JSON Export:

  • Raw data format
  • For developers/integration
  • Includes all metadata
  • Programmatic access

Podcast Exports

ZIP Package:

  • Complete podcast bundle
  • Audio files (WAV format)
  • Transcript with timestamps
  • Metadata and configuration
  • Cover image (if available)

Segments JSON:

  • Developer-friendly format
  • Per-segment timing data
  • Speaker assignments
  • Audio file references

Audio Download:

  • Direct MP3/WAV download
  • Combined final audio
  • Ready for distribution

Advanced Export Options

Customization:

  • Include/exclude timestamps
  • Show/hide citations
  • Custom filenames
  • Cover image upload

Batch Operations:

  • Export multiple chats at once
  • Bulk podcast downloads
  • Archive entire sessions

Troubleshooting

Common Issues and Solutions

PDF Processing Fails:

  • Ensure PDF is not password-protected
  • Check file size (max 200MB)
  • Try with a different PDF to isolate the issue
  • Verify PDF contains extractable text (not just images)

Chat Responses Are Poor:

  • Make sure sources are selected (blue highlighting)
  • Use more specific questions
  • Try rephrasing your question
  • Ensure document processing completed successfully

Podcast Generation Stuck:

  • Check system resources (CPU/memory usage)
  • Verify selected sources are valid
  • Try with fewer or smaller documents
  • Restart the application if needed

Audio Quality Issues:

  • Use higher quality voice settings
  • Ensure good source audio for custom voices
  • Check system audio settings
  • Try different voice combinations

Upload Failures:

  • Check internet connection
  • Verify file isn't corrupted
  • Try uploading a smaller file first
  • Clear browser cache if using web version

Performance Tips

For Better Performance:

  • Close other resource-intensive applications
  • Use shorter documents when possible
  • Generate podcasts during off-peak hours
  • Keep system updated

Managing System Resources:

  • Monitor CPU and memory usage
  • Allow adequate time for processing
  • Use quality settings appropriate for your hardware
  • Consider upgrading hardware for heavy usage

Error Messages

"ML Service Unavailable":

  • MLX models are not loaded
  • Check system compatibility
  • Restart the application
  • Verify installation

"Source Not Found":

  • Source may have been deleted
  • Refresh sources panel
  • Re-upload if necessary

"Generation Failed":

  • Check source selection
  • Verify system resources
  • Try with different content
  • Check error logs

Tips and Best Practices

Getting the Best Results

Document Selection:

  • Use high-quality, well-structured PDFs
  • Academic papers and reports work best
  • Avoid heavily formatted documents
  • Ensure text is machine-readable

Question Asking:

  • Be specific and clear
  • Ask one concept at a time
  • Use proper terminology when available
  • Build on previous questions for context

Podcast Creation:

  • Select complementary documents for richer discussions
  • Use dramatic enhancement for more engaging content
  • Experiment with different voice combinations
  • Consider your target audience

Workflow Optimization

Efficient Source Management:

  • Organize documents by project or topic
  • Use descriptive filenames
  • Remove unused sources regularly
  • Keep source selection relevant to current task

Chat Best Practices:

  • Start with broad questions, then get specific
  • Save important conversations before clearing
  • Use export features to preserve insights
  • Take notes on key findings

Content Creation Strategy:

  • Plan your podcast structure before generation
  • Test voice combinations with short content first
  • Create mind maps to visualize complex topics
  • Use video features to enhance podcast content

Collaboration and Sharing

Sharing Content:

  • Export chats as PDFs for formal sharing
  • Use HTML exports for web publishing
  • Share podcast ZIPs for complete packages
  • Create video content for presentations

Team Workflows:

  • Establish naming conventions for sources
  • Document useful question patterns
  • Share successful voice configurations
  • Create templates for common use cases

Advanced Features

Custom Voice Training

Professional Voice Training:

  1. Record in a quiet environment
  2. Use consistent microphone and settings
  3. Include various speech patterns
  4. Avoid background music or effects
  5. Speak naturally and clearly

Voice Quality Optimization:

  • Use lossless audio formats when possible
  • Normalize audio levels
  • Remove silence at beginning/end
  • Ensure consistent speaking pace

API Integration

For developers interested in integrating NotebookMLX:

  • RESTful API available for all major functions
  • Comprehensive API documentation in /docs/API.md
  • Support for programmatic source management
  • Webhook support for long-running tasks

Customization Options

Interface Customization:

  • Theme selection (light/dark)
  • Panel layout preferences
  • Keyboard shortcuts
  • Export defaults

Model Configuration:

  • Choose different MLX models for various tasks
  • Adjust generation parameters
  • Configure quality vs. speed tradeoffs
  • Set resource limits

Getting Help

Support Resources

Documentation:

  • Complete API documentation
  • Component reference guides
  • Architecture overview
  • Development guides

Community:

  • GitHub repository for issues and feature requests
  • Discussion forums for user questions
  • Example projects and use cases
  • Regular updates and improvements

Technical Support:

  • Error reporting system
  • Diagnostic tools
  • Performance monitoring
  • Update notifications

Remember that NotebookMLX is continuously evolving. Check for updates regularly and consult the latest documentation for new features and improvements.